AI: are analysts no longer needed?
The rapid advances in artificial intelligence are forcing us to think about how technologies and techniques for analytics and decision support are transforming. This vision is essential for anyone developing and applying data governance practices in organizations to understand which ones make sense to invest in. Will familiar tools persist, or will they be completely replaced by new technologies? What place will humans have in these processes?
There are claims that very soon AI algorithms will cope with most analytical tasks better than humans, which will leave many analysts unemployed. Indeed, a machine can very quickly "cover" a much larger amount of information than a human, take into account many factors and produce a result that is most likely to be correct. The emergence of ChatGPT gave the impression of a revolution due to the fact that the machine was able to communicate with a human in a common language: the disappearance of the layer between the algorithm and the decision-maker made the very analysts and programmers who are now the "interface" between business executives and the data processing environment unnecessary.
What can human beings do that AI can't?
Is an AI algorithm really capable of surpassing or replacing humans? This question can be divided into two: whether it is technically possible, and whether it will actually benefit people, organizations, and society.
From a technical point of view, the answer is obvious. Current AI implementations based on machine learning do not generate fundamentally new information: they merely generate an infinite number of variations based on the data that was put into them during training. ChatGPT, one of the most impressive examples of generative neural networks, can "invent a story" or "write a poem" but the results of its work will in any case be a compilation of previously known plot moves, artistic techniques, styles, etc. A person can easily distinguish mass literature from a masterpiece, but ChatGPT is fundamentally unable to create a masterpiece: it is secondary and limited to the reproduction of known samples. The same is true when using it as an analyst: ChatGPT can give trivial recommendations, refer to someone else's experience, but it cannot "invent" new moves in market strategy or innovative product ideas. Meanwhile, it is new and non-standard solutions in business that can give a truly valuable effect and allow you to get ahead of your competitors.
The answer to the question of whether a hypothetical full automation of business decision-making processes could be beneficial is also rather negative. The concept of "Sustainable Growth" appeared precisely because the achievement of quantitatively maximum results - profit growth, production volume, etc. - does not guarantee the approach to important human goals. - does not at all guarantee approaching the goals that are important for a person. Decisions in business, even the most "capitalist", are always made with human values in mind. Obviously, we cannot entrust AI to decide what will be better or worse from a value point of view: this will always be the prerogative of the individual himself.
How will AI change enterprise IT?
This leads us to understand two areas in which AI should not, and in all likelihood will not be able to replace humans in the foreseeable future: creativity and creativity, as well as goal-setting and value judgments. Of course, this also raises the requirements to the analysts themselves: specialists who are not creative and not capable of evaluations will indeed be able to be replaced by an algorithm.
At the same time, AI technologies will undoubtedly change the way corporate analysts work. The next generation of analytical tools will not be limited to information obtained during training and will be able to access external data sources - for example, extracting and processing fresh operational information from business process automation systems. Already today, AI tools are helping to automate the discovery of correlations in data that indicate certain patterns of processes in the real world. It is yet to truly unlock the possibilities of transforming unstructured information (text, graphics) into structured information (sets of formally expressed statements).
It should not be overlooked that today's AI technologies model mostly empirical thinking and pay little attention to simulating conceptual, logical thinking - yet these are the tools that will make AI conclusions explainable and verifiable.
What kind of information does AI work with?
"Hype" AI solutions focus on processing unstructured information - text and graphical (this is also a consequence of modeling primarily empirical thinking). Business applications primarily process structured data, which is always conceptualized, i.e., has formal meaning, can participate in calculations and control checks, etc. The processing of such data by AI tools beyond narrow specific tasks is still less developed. One obstacle is the disparate and (in some cases) difficult interpretation of the meaning of the data.
The transformation of all information available for analysis from a multitude of disparate data sets into a single, coherent representation (Enterprise Knowledge Graph), the availability of unambiguous meaning for each unit of processed information and the ability to determine the level of confidence in it will certainly reduce the share of routine in the work of analysts and increase the space for the manifestation of creative abilities. The application of graph-based analytics algorithms, including machine learning on graphs, will open up new opportunities to extract business value from data. DataVera's solutions enable these benefits now.
We believe that the comprehensive development of data governance practices using AI as one of the tools will bring significant benefits to almost any business. With our vision of technology development, we offer our customers the tools and data practices that will remain relevant in the medium term.
